tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[8,1] = 1000001 is not in [0,

基于TF2.0的keras BiLSTM-CRF模型报错
在加载数据进入模型训练时报如下错误
tensorflow.python.framework.errors_impl.InvalidArgumentError:

tensorflow.python.framework.errors_impl.InvalidArgumentError:  indices[8,1] = 1000001 is not in [0, 1502)
	 [[node sequential/embedding/embedding_lookup (defined at E:\competition\Specification\NerModel.py:60) ]] [Op:__inference_train_function_123177]

Errors may have originated from an input operation.
Input Source operations connected to node sequential/embedding/embedding_lookup:
 sequential/embedding/embedding_lookup/118162 (defined at D:\Anaconda3\envs\tensorflow_2.0\lib\contextlib.py:81)

Function call stack:
train_function

原因：
我的语料字典转换序号时，未知字是用的1000001数字，mapping_dict总长度为1502，而mapping_dict['unk'] = 1000001。1000001超过了其长度因此报错（原理也很奇怪，为什么值不能超过它的长度呢？欢迎留言，感谢！！！）。

解决：
完成保存语料字典时，要将转换的数字排序。即最后语料字典要进行一下操作。

 new_id = 0
    for w in list(mapping_dict):
        mapping_dict[w] = new_id
        new_id += 1